2. Linux.do
  3. Charm Hyper 推出新套餐 $20可用价值$375Token 支持GLM5.2

Charm Hyper 推出新套餐 $20可用价值$375Token 支持GLM5.2

Myth 2026-06-22 04:23 1



https://hyper.charm.land/

支持模型和价格对照表 基本和官方api价格一致


























































































































































































































Lab Model Context $/1M In $/1M Out $/1M Cache Read $/1M Cache Write Features
DeepSeek DeepSeek V4 Flash 1000K $0.1400 $0.2800 $0.0000 $0.0028 Reasoning, Attachments
DeepSeek DeepSeek V4 Pro 1000K $0.4350 $0.8700 $0.0000 $0.0036 Reasoning
Google Gemma 4 26B A4B 256K $0.1020 $0.3940 $0.0000 $0.0510 Reasoning
Z.AI GLM-5 202K $0.9500 $3.0400 $0.0000 $0.4750 Reasoning
Z.AI GLM 5.1 202K $1.4000 $4.4000 $0.0000 $0.2600 Reasoning, Attachments
Z.ai GLM-5.2 1000K $1.5000 $4.5000 $0.0000 $0.3000 Reasoning
OpenAI gpt-oss-120b 131K $0.1780 $0.6800 $0.0000 $0.0890 Reasoning
Moonshot AI Kimi K2.5 262K $0.5370 $2.7150 $0.0000 $0.2685 Reasoning
Moonshot AI Kimi K2.6 262K $0.9500 $4.0000 $0.0000 $0.1600 Reasoning, Attachments
Moonshot AI Kimi K2.7 Code 256K $0.9500 $4.0000 $0.0000 $0.1900 Reasoning, Attachments
Meta Llama 3.3 70B Instruct 128K $0.6066 $1.0386 $0.0000 $0.3033 Reasoning
Meta Llama 4 Maverick 17B 128E Instruct FP8 430K $0.2800 $0.9000 $0.0000 $0.1400 Reasoning
MiniMax MiniMax M2.7 204K $0.3000 $1.2000 $0.3750 $0.0600 Reasoning
Alibaba Qwen3.6-Flash 1000K $1.0000 $4.0000 $0.1000 $1.2500 Reasoning, Attachments
Alibaba Qwen3.6-Max 256K $2.0000 $12.0000 $0.2000 $2.5000 Reasoning
Alibaba Qwen3.6-Plus 1000K $2.0000 $6.0000 $0.2000 $2.5000 Reasoning, Attachments
Alibaba Qwen3.7-Max 1000K $2.5000 $7.5000 $0.2500 $3.1250 Reasoning
Alibaba Qwen3.7-Plus 1000K $1.2000 $4.8000 $0.1200 $1.5000 Reasoning, Attachments
Alibaba Qwen3 Coder 480B A35B Instruct INT4 Mixed AR 106K $0.4450 $2.1450 $0.0000 $0.2225 Reasoning
Alibaba Qwen3 Next 80B A3B Instruct 262K $0.1295 $1.2160 $0.0000 $0.0648 Reasoning

最新回复 (6)
  • 惊鱼 06-22 04:29
    1楼

    挺喜欢 charm 的，以前用他们的库写过 cli，他们出了一个 agent，挺好看，可惜不是很好用

  • a12908 06-22 04:29
    2楼

    谁来尝尝咸淡 高峰期会不会429?实际是否量化?

  • Myth 楼主 06-22 04:32
    3楼

    实测白天deepseek v4 pro 输出150tokens/s 是官方api4-5倍

    晚上反而可能首令延迟比较高

    可能是因为白天没什么国人用 晚上又是国外用户上班

    其他模型消耗太高用的比较少

  • 3GZ 06-22 04:39
    4楼

    不知道这家部署的模型是什么精度的。

  • ninaya 06-22 04:46
    5楼

    这家的glm5.2好像不支持思考，传入思考参数显示错误。。。

  • iceberg01 06-22 04:54
    6楼

    这家是日限啊，付20 刀月套餐，每天 12.5 刀的额度？

* 帖子来源Linux.do
附近帖子
返回
飞读
随机推荐